Autonomous Cars:
Levels of Autonomous Cars:
According to the Society of Automotive Engineers (SAE) cars and vehicles in general are classified into 5 different classifications:                Level 0: Automated system has no vehicle control, but may issue warnings.        Level 1: Driver must be ready to take control at any time. Automated system may include features such as Adaptive Cruise Control (ACC), Parking Assistance with automated steering, and Lane Keeping Assistance (LKA) Type II in any combination.        Level 2: The driver is obliged to detect objects and events and respond if the automated system fails to respond properly. The automated system executes accelerating, braking, and steering. The automated system can deactivate immediately upon takeover by the driver.        Level 3: Within known, limited environments (such as freeways), the driver can safely turn their attention away from driving tasks, but must still be prepared to take control when needed.        Level 4: The automated system can control the vehicle in all but a few environments such as severe weather. The driver must enable the automated system only when it is safe to do so. When enabled, driver attention is not required.        Level 5: Other than setting the destination and starting the system, no human intervention is required. The automatic system can drive to any location where it is legal to drive and make its own decisions.Sensor Technologies:        
Simple cars and other types of vehicles are operated by humans and these humans rely on their senses such as sight and sound to understand their environment and use their cognitive capabilities to make decisions according to these inputs. For autonomous cars and other autonomous vehicles, the human senses must be replaced with electronic sensors and the cognitive capabilities by electronic computing power. The most common sensor technologies are as follows:
LIDAR (Light Detection and Ranging)—is a technology that measures distance by illuminating its surroundings with laser light and receiving the reflections. However, the maximum power of the laser light needs to be kept limited to make them safe for eyes, as the laser light can easily be absorbed by eyes. Such LIDAR systems are usually quite large, expensive and do not blend in well with the overall design of a car/vehicle. The weight of such systems can be as high as tens of kilograms and the cost can be expensive and in some cases high up to $100,000.
Radar (Radio Detection and Ranging)—These days Radar systems can be found as a single chip solution that is lightweight and cost effective. These systems work very well regardless of lighting or weather conditions and have satisfying accuracy in determining the speed of objects around the vehicle. Having said the above, mainly because of phase uncertainties the resolution of Radar systems is usually not sufficient.
Ultrasonic Sensors—These sensors use sound waves and measure their reflections from objects surrounding the vehicle. These sensors are very accurate and work in every type of lighting conditions. Ultrasonic sensors are also small and cheap and work well in almost any kind of weather, but that is because of their very short range of a few meters.
Passive Visual Sensing—This type of sensing uses cameras and image recognition algorithms. This sensor technology has one advantage that none of the previous sensor technologies have colour and contrast recognition. As with any camera based systems, the performance of these systems degrades with bad lighting or adverse weather conditions.
The table below is designed to provide a better understanding of the advantages and disadvantages of the different current sensor technologies and their overall contribution to an autonomous vehicle:
The following tables scores the different sensors on a scale of 1 to 3, where 3 is the best score:
ItemLIDARRADARUltrasonicCameraProximity Detection1231Range2213Resolution2113Operation in darkness3331Operation in light3332Operation in adverse Weather2331Identifies colour or contrast1113Speed measurement2311Size1333Cost1333Total18242221                The above presentation of the state of the technology proved a high-level view of the advantages and disadvantages of the technologies from different perspectives.Drawbacks of Current Sensors:        
As shown in the table above, the available sensors for existing autonomous vehicles are LIDAR, Sonar, passive vision (cameras), and radar. Many of these sensors come with significant drawbacks, while radar systems do not experience most of the drawbacks and thus better among other sensors, based on the table shown above:
For example, LIDAR systems have a “dead zone” in their immediate surroundings (as shown in FIG. 30A), while a Radar system will be able to cover the immediate surroundings of a vehicle as well as long range with enhanced accuracy.
In order to eliminate the “dead zone” as much as possible LIDARs are mounted tall above the vehicle (as shown in FIG. 30B). These limits the options of using parking garages, causes difficulty in the use of roof top accessories and finally also makes the vehicle less marketable since such a tower does not blend in well with the design of a vehicle.
Typical LIDAR systems generate enormous amounts of data which require expensive and complicated computation capabilities, while Radar systems generate only a fraction of this data and reduce the cost and complication of on board computation systems significantly. For example, some types of LIDAR systems generate amounts of 1-Gb/s data that require substantial amount of computation by strong computers to process such high mount of data. In some cases, these massive computations require additional computation and correlation of information from other sensors and sources of information. In some cases, the source for additional computations is based on detailed road information collected over time in databases or in enhanced maps. Computations and correlations can be performed against past information and data collected over time.
Typical LIDAR systems are sensitive to adverse weather such as rain, fog and snow while Radar sensors are not. A radar system will stay reliable and accurate in adverse weather conditions (as shown in FIG. 31). LIDAR systems use mechanical rotation mechanisms that are prone to failure, Radars are solid state and do not have moving parts and as such have a minimal rate of failures.
Typical LIDAR systems rely on a rotation speed of around 5-15 Hz. This means that if a vehicle moves at a speed of 65 mph, the distance the vehicle travels between “looks” is about 10 ft. Radar sensor systems are able to continuously scan their surroundings especially when these systems use one transmitting and one receiving antenna (Bistatic system) (as depicted in FIG. 32). Further, LIDAR systems are not accurate in determining speed and autonomous vehicles rely on Radar for accurate speed detection.
Sonar:
Sonar sensors are very accurate, but can cover only the immediate surroundings of a vehicle, their range is limited to several meters only. The Radar system disclosed in this patent is capable of covering these immediate surroundings as well and with similar accuracy. Further, Sonar sensors cannot be hidden behind cars' plastic parts which poses a design problem, Radars can easily be hidden behind these parts without being noticed.
Passive Visual Sensing (Cameras):
Passive visual sensing uses the available light to determine the surroundings of an autonomous vehicle. In poor lighting condition the performance of passive visual sensing systems degrades significantly and is many times depending on the light that the vehicle itself provides and as such does not provide any benefit over the human eye. Radar systems, on the other hand, are completely agnostic to lighting conditions and perform the same regardless of light (as shown in FIG. 33).
Passive visual sensing is very limited in adverse weather conditions such as heavy rain, fog or snow; Radar systems are much more capable of handling these situations. Passive visual systems create great amounts of data as well which needs to be handled in real time and thus require expensive and complicated computation capabilities, while Radar systems create much less data that is easier to handle. Passive visual systems cannot see “through” objects while Radar can, which is useful in determining if there are hazards behind vegetation for instance such as wildlife that is about to cross the road.
Further, it is easily understandable that in order to cover all possible scenarios all (or most) of these sensors need to work together as a well-tuned orchestra. But even if that is the case, under adverse lighting and weather condition some sensor types suffer from performance degradation while the Radar performance stays practically stable under all of these conditions. The practical conclusion is that Radar performance is not driven by environmental factors as much as by its own technology deficiencies, or a specific deficiency of one of its internal components that the invention here will solve.
Summarizing all of the advantages and disadvantages mentioned above it is clear that Radar systems are efficient in terms of cost, weight, size and computing power. Radar systems are also very reliable under adverse weather conditions and all possible lighting scenarios. Further, SAR Radar Systems may be implemented to create a detailed picture of the surroundings and even distinguish between different types of material.
However, the drawback of existing Radar sensors was the impact on their accuracy due to the phase noise of its frequency source, the synthesizer. Thus, an enhanced system is required (for purposes such as for autonomous vehicles) that may utilize benefits of Radar system by mitigating/eliminating the corresponding existing drawbacks. For example, the required enhanced system, in addition to improving common existing Radar systems, should also improve bistatic or multistatic Radar designs that use the same platform or different platforms to transmit and receive for reducing the phase ambiguity that is created by the distance of the transmitting antenna from the receiving antenna by a significant amount.
Essentially, a signal that is sent out to cover objects (here: Radar Signal) is not completely spectrally clean but sent out accompanied by phase noise in the shape of “skirt” in the frequency domain, and will meet similar one in the receiver signal processing once it is received back. In a very basic target detection systems, fast moving objects will shift the frequency to far enough distance from the carrier so that the weak signal that is being received will be outside of this phase noise “skirt”. Slow moving objects, however, such as cars, pedestrians, bicycles, animals, etc. might create a received signal that is closer to the carrier and weaker than the phase noise and this signal will be buried under this noise and practically will be non-detectable or non-recognizable.
More advanced systems use modulated signals (such as FMCW) but the same challenge to identify slow moving objects remains. The determination of two physically close objects vs. one larger object is also being challenged by phase noise.
Another advanced Radar System worth mentioning is Synthetic aperture Radar (or SAR) that is described in a different section of this disclosure.
Many algorithms and methods have been developed to filter out inaccuracies of Radar based imaging, detection and other result processing. Some are more computational intensive while others are not. The common to all of them is that they are not able to filter out the inherent phase noise of the Radar system itself.
This is crucial since a lot of the information a Radar system relies on, is in the phase of the returning signal. One simple example for this phenomenon is when a Radar signal hits a wall that is perpendicular to the ground or a wall that has an angle that is not 90 degree relative to the surface, the phases of the return signals will be slightly different and this information could be “buried” under the phase noise of the Radar system.
Further, speckle noise is a phenomenon where a received Radar signal includes “granular” noise. Basically, these granular dots are created by the sum of all information that is scattered back from within a “resolution cell”. All of these signals can add up constructively, destructively or cancel each other out. Elaborate filters and methods have been developed, but all of them function better and with less effort when the signals have a better spectral purity, or in other words better phase noise. One of these methods, just as an example, is the “multiple look” method. When implementing this, each “look” happens from a slightly different point so that the backscatter also looks a bit different. These backscatters are then averaged and used for the final imaging. The downside of this is that the more “looks” are taken the more averaging happens and information is lost as with any averaging.
As additional background for this invention there are few phenomena that need to be laid out here:
Doppler Effect:
The Doppler Effect is the change in frequency or wavelength of a wave for an observer moving relative to its source. This is true for sound waves, electromagnetic waves and any other periodic event. Most people know about the Doppler Effect from their own experience when they hear a car that is sounding a siren approaching, passing by and then receding. During the approach the sound waves get “pushed” to a higher frequency and thus the siren seems to have a higher pitch, and when the vehicle gains distance this pitch gets lower since the sound frequency is being “pushed” to a lower frequency.
The physical and mathematical model of this phenomena is described in the following formula:
  f  =            (                        c          +                      v            r                                    c          +                      v            s                              )        ⁢          f      0      Where f0 is the center frequency of the signal, c is the speed of light, vr is the velocity of the receiver relative to the sound/radiation source, vs is the velocity of the sound source relative to the receiver and f is the frequency shift that is being created.After simplifying the equation, we will get:
      Δ    ⁢                  ⁢    f    =                    Δ        ⁢                                  ⁢        v            c        ⁢          f      0      Where Δv is the relative velocity of the sound source to the receiver and Δf is the frequency shift created by the velocity difference. It can easily be seen that when the velocity is positive (the objects get closer to each other) the frequency shift will be up. When the relative velocity is 0, there will be no frequency shift at all, and when the relative velocity is negative (the objects gain a distance from one another) the frequency shift is down.
In old fashioned Radars the Doppler effect gets a little more complicated since a Radar is sending out a signal and expects to a receive signal that is lower in power but at the same frequency when it hits an object. If this object is moving, then this received signal will be subject to the Doppler effect and in reality, the received signal will not be received at the same frequency as the frequency of the transmit signal. The challenge here is that these frequency errors can be very subtle and could be obscured by the phase noise of the system (as shown in FIG. 34). The obvious drawback is that vital information about the velocity of an object gets lost only because of phase noise (see figure below). The above is especially right when dealing with objects that move slower than airplanes and missiles, such as cars, bicycles, pedestrians, etc.
Modulated Signals—Newer Radar systems use modulated signals that are broadly called FMCW (Frequency Modulated Continuous wave), but they can come in all forms and shapes such as NLCW, PMCW, chirps, etc. (Nonlinear Continuous Wave and Phase Modulated Continuous Wave). The main reason for the use of modulated signals is that old fashioned Radars need to transmit a lot of power to receive and echo back from a target while modulated signals and smart receive techniques can do that with much lower transmit power.
Another big advantage of FMCW based Radar systems is that the distance of a target can be calculated based on Δf from the instantaneous carrier signal rather than travel time. However, herein also lies the problem—to be able to calculate and determine the characteristics of a target accurately a spectrally clean signal with ultra-low phase noise as low as technically possible provides many advantages.
Usually modulated Radar signals are processed with the help of FFT utilizing signal processing windows and pulse compression algorithms. While these methods are good phase noise still remains one of the major contributors, if not to say the largest contributor to errors and inaccuracies.
The spectral picture of a processed signal looks like the FIG. 35. As one can see the spectral picture contains also unwanted sidelobes. One major contributor to the sidelobes is the phase noise of the Radar system. Phase noise (or sometimes also called Phase Jitter or simply Jitter) responds to 20 log(integrated phase noise in rad). This spectral regrowth of side lobes can cause errors in the determination of the actual distance of a target, and can obscure a small target that is close to a larger target. It can also cause errors in target velocity estimation.
Another use of Radar that is sensitive to phase noise includes Synthetic Aperture Radar (SAR) of all kinds. These Radars are being used in countless applications ranging from space exploration through earth's surface mapping, Ice pack measuring, forest coverage, various military applications to urban imaging and archaeological surveys. However, all the Radar applications have a common drawback of bearing phase noise that leads to depletion of the quality of the end-result or failure in achieving the desired outcome. For example, whether we refer to Interferometric SAR (InSAR) or Polarimetric SAR (PolSAR) or a combination of these methods or any other type of SAR or Radar in general, all of them are suspect to phase noise effects regardless of the type of waveform/chirp used. Considering the shift in frequency and the low signal strength there is a probability that the received Radar signal will be buried under the phase noise skirt, and the slower the object this probability grows. Again, the determination of two close objects vs. one large one is a challenge here. SAR Radars create images of their surroundings and the accuracy of the images depends also on the phase noise of the signal. Some of these radars can also determine electromagnetic characteristics of their target such as the dielectric constant, loss tangent etc. The accuracy here again depends on the signal quality which is largely determined by the sidelobes created during the utilization of the FFT algorithm mentioned above which in turn stem from the phase noise of the system.
Further, in FIG. 36, the sidelobes have been simplified to only a wide overlapping area. This is also very close to what happens in reality because of the way the signal processing algorithms work. As can be easily seen in the figure above, weaker return signals can get obscured in the sidelobes of a stronger signal and the overall available and crucial SNR decreases because every return signal carries sidelobes with it.
Thus, the Radar systems are challenged when dealing with slow moving objects such as cars, bicycles and pedestrians. Furthermore, these traditional Radar systems, whether using a modulated or non-modulated signal, have difficulties identifying objects that are very close to each other since one of them will be obscured by the phase noise of the system.
Based on the aforementioned, there is a need of a radar system that can effectively be utilized for autonomous cars by canceling or reducing the phase noise of the received Radar signal. For example, the system should be capable of determining surroundings (such as by detecting objects therein) with cost effectiveness and without significant internal phase noise affecting performance.
The system should be able to add as less as possible phase noise to the received Radar signal. Further, the system should be able of detecting and analyzing the received signal without being affected by the internal receiver phase noise therefrom. Furthermore, the system should be capable of implementing artificial intelligence to make smart decisions based on the determined surrounding information. Additionally, the system should be capable to overcome the shortcomings of the existing systems and technologies.